P Olyanhydride Degradation and Erosion A

Advanced Drug Delivery Reviews 54 (2002) 911–931 www.elsevier.com/locate/drugdeliv

P olyanhydride degradation and erosion A. Gopferich¨ * , J. Tessmar

Faculty of Pharmacy and Chemistry, Pharmaceutical Technology Unit, University of Regensburg, Universitatsstrasse¨ 31, D-93053 Regensburg, Germany Received 17 March 2002; accepted 19 June 2002

Abstract

It was the intention of this paper to give a survey on the degradation and erosion of polyanhydrides. Due to the multitude of polymers that have been synthesized in this class of material in recent years, it was not possible to discuss all polyanhydrides that have gained in significance based on their application. It was rather the intention to provide a broad picture on polyanhydride degradation and erosion based on the knowledge that we have from those polymers that have been intensively investigated. To reach this goal this review contains several sections. First, the foundation for an understanding of the nomenclature are laid by defining degradation and erosion which was deemed necessary because many different definitions exist in the current literature. Next, the properties of major classes of anhydrides are reviewed and the impact of geometry on degradation and erosion is discussed. A complicated issue is the control of drug release from degradable polymers. Therefore, the aspect of erosion-controlled release and drug stability inside polyanhydrides are discussed. Towards the end of the paper models are briefly reviewed that describe the erosion of polyanhydrides. Empirical models as well as Monte-Carlo-based approaches are described. Finally it is outlined how theoretical models can help to answer the question why polyanhydrides are surface eroding. A look at the microstructure and the results from these models lead to the conclusion that polyanhydrides are surface eroding due to their fast degradation. However they switch to bulk erosion once the device dimensions drop below a critical limit.  2002 Elsevier Science B.V. All rights reserved.

Keywords: Bioerosion; Polyanhydride; Polymer degradation; Polymer erosion; Modeling

Contents

1 . Introduction ...... 912 2 . Polymer degradation and erosion...... 913 2 .1. Surface erosion versus bulk erosion...... 913 2 .2. Physicochemical characterization of polyanhydride degradation and erosion...... 914 3 . Parameters affecting polyanhydride degradation and erosion ...... 915 3 .1. The impact of polymer composition on degradation and erosion ...... 915 3 .1.1. Aliphatic polyanhydrides ...... 915 3 .1.2. Aromatic polyanhydrides ...... 916

*Corresponding author. Tel.: 149-941-943-4843; fax: 149-941-943-4807. E-mail address: [email protected] (A. Gopferich).¨

3 .1.3. Polyanhydride copolymers derived from aromatic and aliphatic monomers...... 917 3 .1.4. Polyanhydrides derived from fatty acids...... 921 3 .1.5. Cross-linked polyanhydrides ...... 923 3 .2. The impact of geometry on degradation and erosion ...... 923 3 .2.1. Macroscopic matrices ...... 923 3 .2.2. Microparticles...... 924 4 . Erosion-controlled drug release from polyanhydrides ...... 924 5 . Polyanhydride erosion modeling...... 925 5 .1. Empirical models ...... 925 5 .2. Monte Carlo-based models ...... 925 6 . Why polyanhydrides undergo surface erosion...... 926 7 . Summary and outlook...... 927 References ...... 928

1 . Introduction control degradation and, most importantly, they are approved as biocompatible which is a tremendous Degradable polymers have attracted significant advantage over new degradable polymers that have attention for use in numerous medical and biomedi- to undergo time- and cost-intensive biocompatibility cal applications that require the presence of a testing. There have been only a few cases in recent material only for a limited period of time [1]. years, in which new degradable polymers were Especially after implantation into the body, it is custom made for application in humans: poly- highly desirable that the material ‘disappears’ to anhydrides are one of these (Fig. 1). obviate the need for any post-application removal. Polyanhydrides were made with the intention to Many current concepts in the pharmaceutical and have a material at hand that fits to a paradigm as old biotechnological field depend significantly on this as biodegradable materials themselves: the material strategy [2,3]. A good example is parenteral drug should degrade within the time frame of their delivery, by which a dose of drug is typically application. For degradables used in controlled re- intended to be released over an extended period of lease applications this means that the completeness time [4]. Biodegradable polymers can help signifi- of polymer erosion coincides with the end of drug cantly to overcome numerous problems inherent to release. This is hard to achieve with polymers that this concept. The polymers can stabilize the drug degrade over weeks such as PLA and PLGA when reservoir from premature inactivation; concomitant- the drug is intended to be released for only a few ly, the polymer can control the release of drug out of days. Therefore in the early 1980s polyanhydrides the reservoir and finally the degradability of the were discovered for drug delivery applications [6]. material helps to overcome the need for any post- The advantage of polyanhydrides is that they are application removal. made of the most reactive functional group available Based on the outlined advantage of degradable for degradation on the base of passive hydrolysis. polymers there have been numerous polymers ex- How this translates to an enhanced degradation and plored for their suitability to degrade in a biological in a further step to an accelerated drug release was environment. Since the 1970s a plethora of materials subject to a careful characterization of poly- has been synthesized. Moreover, strategies are under anhydrides in the following decades. way to provide polymers on the basis of com- binatorial approaches [5]. Surprisingly, the class of hydrophobic biodegradable polymers has been domi- nated by poly(a-hydroxy acids) such as poly(lactic acid) (PLA) and poly(lactic-co-glycolic acid) (PLGA) for more than 30 years. These materials are available in different compositions that allow to Fig. 1. General polyanhydride structure. A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931 913

It is the intention of this review to shed some light achieve with our experiments is to find at least a on polyanhydride degradation and erosion in vitro to crude forecast of the in vivo degradation behavior. elucidate the consequences that result from their That this is not a trivial task is documented by degradation and erosion behavior. First, both pro- numerous publications on this issue in the current cesses will be defined and methods will be presented literature. Polyanhydrides such as poly[1,3-bis( p- by which we can investigate them. The erosion carboxyphenoxy) propane-co-sebacic acid] (p(CPP- mechanism of some classes of polyanhydrides will SA)), for example, were found to erode substantially be reviewed and finally theoretical models will be slower in vitro than in vivo [11]. Our efforts in vitro discussed that describe polyanhydride erosion. should, therefore, focus primarily on standardizing the experimental conditions rather than trying (usually in vain) to mimic in vivo conditions. The latter is 2 . Polymer degradation and erosion usually only successful for a few materials and fails whenever we try to expand on a larger number of Degradation in this review designates the process polymers. For purposes of comparing results we of polymer chain cleavage, a definition that was also usually perform erosion experiments in buffer solu- adapted for polyanhydrides [7]. The prefix ‘bio’, tions. Many experiments are carried out in phosphate thereby, usually indicates that in a biological system buffer solution of pH 7.4 at 37 8C. When erosion data there are other mechanisms besides passive hydrol- are discussed below these will be the experimental ysis that contribute to the kinetics of the process. conditions if not stated otherwise. Such processes that are usually mediated by an enzymatic mechanism have been shown to affect the hydrolysis of poly(a-hydroxy acids) in vivo [8]. 2 .1. Surface erosion versus bulk erosion Whether they are of any enhancing effect for fast- degrading polymers such as polyanhydrides remains, The fast degradation of polyanhydrides based on however, questionable [9]. As polyanhydrides belong their chemical structure has tremendous conse- to the class of water-insoluble hydrophobic poly- quences for their erosion properties. In contrast to mers, it is mandatory for these materials to degrade PLA and PLGA, polyanhydrides are undergoing prior to erosion. Erosion in this review designates surface erosion which is also termed heterogeneous the sum of all processes that can lead to the loss of erosion. The mechanism is schematically illustrated mass from a polyanhydride matrix irrespective of its in Fig. 2. Typically, degradation and erosion are geometry, such as slab, cylinder or microsphere. It is limited to the surface of a polymer only (Fig. 2a). In obvious that polyanhydrides need to undergo degra- an ideal scenario, the mass loss kinetics are, there- dation prior to erosion. However, it must always be fore, linear. When degradation is the only mechanism borne in mind that processes other than degradation that controls the erosion process the molecular can contribute to erosion as well. It has, for example, weight of the polymer should be constant. Bulk been shown for some polyanhydrides that cracks erosion also termed homogeneous erosion, in con- form early during degradation on the surface of trast, reflects a different mechanism. Bulk eroding polymers matrix discs [10], which might lead to the polymers degrade all over their cross-section and loss of pieces of non-degraded material due to have erosion kinetics that are non-linear and are mechanical instabilities. Furthermore some poly- usually characterized by a discontinuity (Fig. 2b) anhydride matrices turn into fragile and brittle [12]. materials so that parts of the matrix may wear off Polyanhydrides usually undergo a linear mass loss under the weak mechanical forces that are applied during erosion, causing us to classify them as surface during in vitro erosion experiments. eroding. However, degradation is not strictly limited Those who are active in the field of degradable only to the surface of a polyanhydride matrix. This is polymers are familiar with the question for the illustrated by numerous experiments in which the optimal conditions that should be used to investigate molecular weight was found to drop exponentially polymer erosion in vitro. What we usually want to [13,14]. 914 A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931

polymer particles [16]. The price of studying degradation in polymer bulk is that we cannot assume that at every point of the matrix identical conditions prevail. That this may have a tremendous impact was shown by investigating different zones of eroding polyanhydride matrices. Major differences existed between polymer areas on the surface and in the matrix center [7,17]. The methods that were used to investigate degradation of polyanhydride matrices are the classical ones used to determine polymer molecular weight. For reasons of simplicity and of automatization, gel permeation chromatography has been used extensive- ly [10,13]. For some polyanhydrides such as p(CPP- SA), data on intrinsic viscosities may even be found [18], so that a universal calibration [19] may be possible. Other methods that have been used to investigate polyanhydride degradation are: infrared spectroscopy by which the anhydride and carbonyl signals of carboxylic acids can be monitored [20],1 H NMR [21],13 C magnetic angle spinning solid-state NMR [16] and differential scanning calorimetry (DSC) [22]. Fig. 2. Schematic illustration of surface erosion and bulk erosion. While the degradation of polyanhydrides can be unequivocally described on the basis of decreasing molecular weight of the polymer, erosion cannot be assessed based on a single parameter. By far the most important measure for polymer erosion is the 2 .2. Physicochemical characterization of mass loss of a polymer device [10]. Based on the polyanhydride degradation and erosion definition of surface erosion one would expect, for a matrix with predominantly slab geometry, a linear The degradation of polymers in general, as well as mass loss profile. It is, however, important to realize the degradation of polyanhydrides, is not easy to that mass loss does not reveal anything of the investigate. What we are usually interested in is the mechanism of erosion. Techniques using spatial degradation kinetics of individual bonds inside a resolution, that allow further investigation into ero- polymer chain. As polyanhydrides are, however, sion mechanisms are: light microscopy techniques essentially insoluble in water we, therefore, need [17], scanning electron microscopy (SEM) [10], organic solvents that contain at least traces of water scanning confocal microscopy [10], atomic force to achieve this goal [15]. Although it may be microscopy [23,24] and surface plasmon resonance possible to compare different polymers for their [24,25]. To investigate the composition of the poly- degradation rate under these well-defined conditions, mer bulk under erosion, a number of physicochemi- it may not allow to predict the overall degradation of cal techniques have been applied, such as differential the chains once they are embedded in a polymer scanning calorimetry [10,22,26], wide-angle X-ray matrix. Therefore, polyanhydride degradation and diffraction [10,26], nuclear magnetic resonance im- erosion have been investigated by incubation of aging (MRI) [27], electron paramagnetic resonance matrices in buffer solutions. In rare cases, water (EPR) [28], following the release of monomers or vapor also has been used to study the degradation of model compounds [29,30] and others. A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931 915

3 . Parameters affecting polyanhydride measuring melting endotherms by DSC during ero- degradation and erosion sion [10]. The crystalline polymer areas are embedded in spherulites, the structure of which becomes 3 .1. The impact of polymer composition on more apparent in AFM images after the surface has degradation and erosion undergone erosion [23]. AFM images taken in situ provided additional evidence that amorphous poly- It is almost impossible to review the erosion mer areas erode faster than crystalline ones. The behavior of all classes of polyanhydrides. In recent erosion profile of p(SA) in phosphate buffer, pH 7.4, years, especially, the number of polyanhydrides has is linear, as expected from a surface eroding polymer increased tremendously. While during the early years (Fig. 4) [10]. of their development they were mainly manufactured Some authors have described the properties of from linear diacids [31–33], they are now increas- other aliphatic polyanhydrides, such as those made ingly made of materials that deviate from that of adipic, pimelic, suberic, azelaic, dodecanedioic scheme. Examples are poly(anhydrides-co-imides) and dodecanedicarboxylic acid [9]. All of these that hold promise for the development of vaccines polymers were rigid, crystalline materials with melt- [34–37], polyanhydride prodrugs [38–40], anhy- ing points increasing with the monomer chain length. drides that carry other degradable bonds in their When rectangular matrices (3 3 7 3 11 mm) were backbone [41], block copolymers with poly(ethylene eroded in vitro, polyanhydrides made of long chain glycol) [42] and many others. Rather than reviewing monomers (7–10 methylene units) lost 20% while all these materials the focus will be on major classes those made of short chain monomers (4–6 methylene of materials for which a decent amount of degra- units) lost 70% of mass during the first 48 h. The dation and erosion data exists and from which example illustrates that aliphatic polyanhydrides general rules of the two processes may be deducted. usually erode fast and are, therefore, not unequivo- Fig. 3 lists some of the monomers that have been cally useful for biomedical and pharmaceutical appli- used for the manufacture of polyanhydrides. As the cations. IUPAC names of monomers and polymers would Some aliphatic polyanhydrides serve special appli- sometimes be rather lengthy the abbreviations given cations such as p(FA-SA), i.e., copolymers derived in parentheses in Fig. 3 will be used to abbreviate from sebacic and fumaric acid (FA). p(FA-SA) monomer and polymer names. A copolymer made of polymers, for example, were proposed for the de- 1,3-bis-( p-carboxyphenoxy)propane (CPP) and velopment of bioadhesive materials that interact with sebacic acid (SA), for example, will be abbreviated mucosal tissues [43]. When microspheres made of p(CPP-SA). The monomer ratio after the abbrevia- p(FA-SA) 20:80, p(FA-SA) 50:50 and p(FA-SA) tion of copolymers will indicate the percent (w/w) 70:30 were degraded in vitro at pH 4.2, 7.4 and 8.8, share of the monomers. the degradation rate at pH 8.8 was significantly higher than at neutral or acidic pH [20]. However, despite rapid degradation, long-lasting anhydride 3 .1.1. Aliphatic polyanhydrides oligomers seemed to persist for extended periods of Aliphatic polyanhydrides were among the first time. When p(FA-SA) 20:80 microspheres were materials to be investigated for the purpose of drug loaded with 2% bovine serum albumin, the degra- delivery. Homopolymers are often problematic ma- dation of the polymer was accelerated [22], indicat- terials as they are usually highly crystalline with ing that additives such as drugs can have a consider- unfavorable mechanical properties. p(SA) for exam- able effect on degradation and erosion. ple has a crystallinity of 66% [26]. Its microstructure In some cases, aliphatic polyanhydrides have been is composed of crystalline and amorphous domains used for blending other polymers. As a fast degrad- which is of utmost importance for the erosion ing component, they are degradation accelerators for mechanism. Erosion preferentially affects the amor- polymers by lowering pH during erosion. Blending phous parts of p(SA), which has been proven by poly(trimethylene carbonate) (PTMC) with poly- 916 A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931

Fig. 3. Monomers that have been used for the synthesis of polyanhydrides.

(adipic anhydride) p(AA), for example, allowed 3 .1.2. Aromatic polyanhydrides control of the erosion of the blend. While matrices The development of aromatic polyanhydrides was containing 20% p(AA) lost only approximately 18% slow over the last decades. One of the reasons is weight over 25 days in PBS at pH 7.4, the mass of certainly their low degradation rates and their hydro- matrices containing 80% p(AA) dropped to approxi- phobic nature, which are disadvantageous for paren- mately 25% of the original value [44]. While the teral applications. p(CPP), for example, is an ex- enhanced mass loss may be mainly attributed to the tremely slow eroding material [45] that has a very polyanhydride component, an accelerated degrada- high melting point of approximately 240 8C [33], so tion of PTMC was caused by presence of the that it is not readily processable. Furthermore, its anhydride. solubility in organic solvents is poor. A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931 917

Fig. 3. (continued)

Materials that seem to overcome some of these 3 .1.3. Polyanhydride copolymers derived from drawbacks are aromatic, ortho-substituted poly- aromatic and aliphatic monomers anhydrides such as poly[1,6-bis(o-carboxyphenox- The examples above make it obvious that, from a y)hexane] p(o-CPH) [46]. Compression-molded 13- historical perspective, the properties of poly- mm diameter discs possessed elastic properties and anhydrides needed to be improved in order to obtain degraded according to ﬁrst-order kinetics. They materials with superior properties, i.e., better me- eroded in a linear way over a time period of chanical characteristics and materials with adjustable approximately 17 days, during which p(CPP) would erosion times. There have been numerous approaches erode only to less than 5% [45]. to do so [33]; however, one of the most successful 918 A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931

Fig. 3. (continued) polymer types were copolymers made of sebacic acid When the degradation of p(CPP-SA) matrices (cylin- and 1,3-bis( p-carboxyphenoxy)propane (p(CPP- drical, 8 mm diameter/1.6 mm height) was investi- SA)). First reports of using these polymers as a gated, it was observed that the molecular weight biomaterial go back to the early 1980s [6]. Since decreased according to first-order kinetics [14]. then, there have been numerous reports on their While varying the molecular weight between 10 and synthesis [15,33,47,48] and characterization [18]. 65 kDa had no significant impact on the degradation This made p(CPP-SA) [16,18,26,27,32,45,49–51] kinetics, it was found that the erosion was delayed in probably the best characterized material in the family a linear manner by 8–12 min. These results may be of polyanhydrides. This development was certainly surprising as one would assume that polyanhydrides spurred by the use of p(CPP-SA) 20:80 for the are surface eroding. However, experiments with development of gliadel [52,53], a BCNU-loaded drug polyanhydrides derived from fatty acid dimers delivery system for the local therapy of malignant (p(FAD-SA)), revealed a significant water uptake, a glioma [54,55]. For p(CPP-SA), a number of fun- prerequisite to polymer degradation [56]. Once a damental erosion and degradation data was raised. polyanhydride matrix has taken up considerable A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931 919

Fig. 3. (continued) amount of water it is highly likely that all polymer lated to other aliphatic/aromatic polyanhydrides is parts in contact with it start degrading. That the that anhydride bonds between aliphatic carboxylic degradation rate is a function of the polymer struc- acids degrade faster than those between aromatic ture and also that of the the monomer, has been ones. This was, for example, proven by using FTIR systematically studied by investigating homologous spectroscopy to follow the intensity of the anhydride series of poly[bis( p-carboxyphenoxy)alkanes]. In- bands that stemmed from SA-SA and CPP-CPP creasing the number of methylene groups from one bonds [7]. These results were later confirmed by1 H to six decreased the degradation rate by three orders NMR [21] and solid state NMR [16]. These differ- of magnitude [45]. A general relationship that was ences in reactivity were finally also confirmed with derived from p(CPP-SA) and that may be extrapo- other polyanhydrides. In degradation experiments with poly(v-( p-carboxyphenoxy) alkanoic anhydrides) derived from v-( p-carboxyphenoxy)acetic acid, -valeric acid and -octanoic acid (CPA, CPV and CPO) [15] anhydride bonds between two aliphatic chains were cleaved faster than between bonds linking two aromatic ends. That the degradation of polyanhydrides is pH dependent was impressively shown with p(CPP) matrix discs eroded in phosphate buffer in the range of pH 7.4–10. Obviously the polymers were most stable at low pH (Fig. 5) [45]. In contrast to degradation, the erosion of p(CPP- SA) is substantially more complicated. As outlined above, erosion lags behind degradation by a couple of minutes [14]. The first changes that can be observed affect the polymer surface. By light microscopy, cracks became visible on the surface of Fig. 4. In vitro erosion profiles of polyanhydride matrix cylinders. p(CPP-SA) 20:80 matrices upon contact with water The cylindrical matrices (14 mm diameter/1 mm height) were eroded in phosphate buffer, pH 7.4, at 37 8C. Reproduced with [10], which was also observed for p(FAD-SA) 50:50 permission from Ref. [10]. (a) p(SA), (b) p(CPP-SA) 20:80, (c) matrices [56]. An observation that is unique for p(CPP-SA) 50:50. polyanhydrides, and that was reported quite early, is 920 A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931

Fig. 5. Erosion of p(CPP) matrix discs as a function of pH. The cylindric matrices (14 mm diameter/1 mm height) were eroded in phosphate buffer, pH 7.4, at 37 8C. Reproduced with permission from Ref. [45]. their almost linear erosion kinetics. Polymers such as p(SA), p(CPP-SA) 20:80, p(CPP-SA) 50:50, p(CPP- SA) 85:15 and p(CPP) were reported to erode at an almost constant rate (Fig. 4) [10,45]. Due to the slower degradation of aromatic anhydride bonds, the erosion velocity decreases with increasing CPP content. A feature that is unique for p(CPP-SA) and many other polyanhydrides is the creation of erosion zones in which the polymer is undergoing erosion [7,29]. Erosion zones in p(CPP-SA) are separated from non-eroded polymer by erosion fronts, which move at constant velocity from the surface of a matrix into its center [10]. While the erosion zone was highly porous, the bulk showed no porosity at all. In the case of cylindrical p(CPP-SA) 20:80 and p(CPP-SA) 50:50 matrix discs (1.4 mm diameter, 1 Fig. 6. Monomer release from polyanhydrides. The cylindric mm height), 70 and 50% mass were lost after 6 days, matrices (14 mm diameter/1 mm height) were eroded in phos- while the overall geometry did not change [10]. This phate buffer, pH 7.4, at 37 8C. Reproduced with permission from Ref. [10]. (A) p(SA), (B) p(CPP-SA) 20:80, (C) p(CPP-SA) erosion behavior is based on the polymer micro- 50:50. structure which is composed of amorphous and crystalline polymer parts, of which the crystalline SA) 50:50 and p(CPP-SA) 20:80 matrices was structures were significantly more erosion resistant similar to that from p(SA) and significantly faster than amorphous ones, and maintained a highly than CPP release [10]. Fig. 6 reveals that the release porous polymer skeleton in the erosion zone over a velocity of CPP increases significantly once SA substantial period of time. Besides following the release is complete. It was hypothesized that these remaining mass of a matrix to assess erosion, the results are due to the solubility of monomers which release of degradation products has been used exten- is different for SA and CPP [10]. Both monomers are sively to follow erosion. This revealed very complex protolytes and, therefore, decrease the pH inside the erosion patterns in the case of p(CPP-SA) [7] (Fig. porous erosion zone. EPR imaging of p(CPP-SA) 6). It was found that the release of SA from p(CPP- 20:80 matrix discs eroded in vitro showed that a pH A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931 921

Fig. 7. pH profiles inside the erosion zone of p(CPP-SA) 20:80 (Mwaa45 000) matrix discs. pK designates the pK of the pH sensitive spin probe. Reproduced with permission from Ref. [28]. gradient with a pH of 7.4 on the matrix surface to resistance of polyanhydrides and concomitantly slow approximately 4.7 at the erosion front develops with down the intrusion of water into polymer matrices. time (Fig. 7) [28]. The pH inside the erosion zone One of the first materials to be synthesized and by seems, thereby, to be mainly controlled by SA that far the best investigated are copolymers made of has an approximately five times higher solubility sebacic acid and a erucic acid (fatty acid) dimer compared to CPP [10]. This suppresses the solubility (FAD) [58]. When rectangular slabs (200 mg; 3 3 and, therefore, the flux of CPP as long as significant 5 3 10 mm) of the material were degraded in vitro, amounts of SA reside in the matrix. After 7 days, their molecular weight dropped again exponentially, however, when SA release is almost complete, the similar to the degradation pattern of p(CPP-SA) [58]. solubility and release of CPP increases significantly. Cylindrical discs (14 mm diameter/0.1 mm height)

This is also reflected by the monomer content of made of p(FAD-SA) 20:80 (Mw 30 000), p(FAD-SA) matrices during erosion (Fig. 8). The monomer 50:50 (Mww25 000) and p(FAD-SA) 70:30 (M content of p(CPP-SA) 20:80 after 4 days reaches a 50 000) all had a molecular weight of less than 5000 maximum of approximately 35%. This is more than after 24 h erosion in vitro [17]. Despite its unique the pores can dissolve. Therefore, one can assume chemical composition with long aliphatic chains, that a good part of the monomer will exist in a p(FAD-SA) erosion behavior resembles that of crystalline state inside the erosion zone. This was p(CPP-SA) in many ways. Polymers with an SA supported by a diffusion/erosion models that were content of approximately 25% and more were found developed to describe the release of monomers from to be semi-crystalline [17,58]. The microstructure of p(CPP-SA) matrices [57]. crystallites revealed a spherulitic arrangement as in p(CPP-SA) [57]. The polymers also formed erosion 3 .1.4. Polyanhydrides derived from fatty acids zones during erosion. Due to the low solubility of The basic idea behind developing polyanhydrides FAD it consisted mainly of a semisolid mixture of from fatty acids was to obtain materials with a FAD and FAD salts [57]. The semisolid layer can pronounced hydrophobic character and improved form a permeation barrier with significant influence mechanical properties compared to materials such as on the release of SA as well as drugs [27]. Acid p(CPP-SA). The hope was to increase the hydrolytic orange, for example, a hydrophilic dye, was released 922 A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931

Fig. 8. Monomer content of polyanhydride matrix cylinders during erosion. The cylindrical matrices (14 mm diameter/1 mm height) were eroded in phosphate buffer, pH 7.4, at 37 8C. Reproduced with permission from Ref. [10]. (A) p(SA), (B) p(CPP-SA) 20:80, (C) p(CPP-SA) 50:50. more slowly in vitro the higher the FAD content of gel-like diffusion barriers, this might explain why the the investigated p(FAD-SA) was [59]. The inves- release of SA was the lowest for polymers made of tigation of eroding p(FAD-SA) 50:50 and p(FAD- equal amounts FAD and SA [60]. SA) 20:80 matrices by MRI and ESR shed some More recently, polyanhydrides were developed light on the changes of the polymer microstructure from non-linear hydrophobic fatty acid esters [61]. during the process [27]. When composed mainly of The development of these polymers was spurred by SA, the polymer was able to maintain a porous the long in vivo half-lives of FAD as well as the erosion zone, while at an equal amount of SA and observation that matrices of p(FAD-SA) 20:80, for FAD the erosion zone became more semisolid. example, underwent signiﬁcant tissue encapsulation Sebacic acid was found again to precipitate inside in rats 65 days post implantation [27]. The polymers the erosion zone. Together with the formation of were synthesized from ricinoleic acid maleate A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931 923

(RAM), 12-hydroxystearic acid succinate (HSAS) then photopolymerized using UV light and appro- and 12-hydroxystearic acid maleate (HSAM) [61]. priate photo initiators [71]. When discs of 16 mm Copolymers, such as p(RAM-SA) 50:50, p(HSAM- diameter and 1.6 mm height were eroded in vitro, SA) 50:50 and p(HSAS-SA) 50:50, were reported to they showed linear erosion profiles. With increasing undergo a sharp decrease of molecular weight during MCPH content, erosion slowed down significantly. the first 24 h of erosion in vitro and lost 40% of their While p(MSA) eroded completely in a few hours, anhydride bonds in 48 h, as determined by FTIR p(MCPH) lost only approximately 20% weight in 80 [61,62]. Their erosion profiles were all similar with a days. Copolymers derived from MSA and MCPH weight loss of 40% after 2 days [61]. showed erosion behavior between these extremes Polyanhydrides were also made of pure fatty acids [71]. that were not modified to carry a second carboxylic acid end [63,64]. These polymers were essentially 3 .2. The impact of geometry on degradation and polyanhydrides made of p(SA) chains terminated by erosion fatty acids such as octanoic (OCTA), lauric (LAUA), myristic (MYA), oleic (OLA) or stearic acid (STA). Most of the information that was collected on the When 300-mg slabs (3 3 7 3 11 mm) were eroded in degradation and erosion of polyanhydrides was vitro the degradation of all copolymers with a collected from the investigation of macroscopic composition of 10 or 30% fatty acid and 90 or 70% matrices. Doing so has the advantage that the sebacic acid, respectively, again showed an exponen- polymer samples can easily be characterized. How- tial loss of molecular weight. The same results were ever, the use of polyanhydrides for drug delivery obtained for non-linear fatty acid anhydrides that applications led to extensive research efforts to were composed of sebacic acid and ricinoleic acid develop and investigate polyanhydride microspheres esterified with C8–C18 fatty acids at their alcohol [72–77]. In the case of monolithic matrices the effect function [64]. The erosion profiles of these polymers of geometry was also investigated. are intriguing. The mass of matrices under erosion remains stable for a couple of days before mass loss 3 .2.1. Macroscopic matrices sets in [63]. This is usually typical of bulk eroding The interest in investigating the effect of geometry polymers. The more fatty acid these polymers con- of monolithic matrices on the erosion of poly- tain, and the longer their chain length, the more anhydrides was spurred by the desire to better pronounced this effect is. It seems, therefore, likely control the release of drugs from implants such as that the solubility of the fatty acids has a pronounced gliadel. Given that a polyanhydride undergoes per- impact on erosion. fect surface erosion it is obvious that the geometry of a device can significantly affect the release kinetics. 3 .1.5. Cross-linked polyanhydrides A few studies focused on this issue. When for The often limited mechanical stability of poly- example cylindrical matrices (5 mm diameter/0.5 anhydrides has always been a handicap to their use mm height, 9 mm diameter/0.8 mm height and 12.5 as biomaterials for orthopedic applications such as a mm diameter/1.4 mm height) were prepared from temporary replacement in bone defects. To overcome p(CPP-SA) 40:60, the amount of water uptake of these limitations unsaturated polyanhydrides that matrices in vitro was a function of size [13]. allow for cross-linking, such as p(FA) and p(FA- Concomitantly the molecular weight of matrices SA), have been synthesized [65]. In recent years dropped in the usual exponential way, however, the unsaturated polyanhydrides were studied and de- rate for bigger matrices was lower than that for veloped more intensively [66–71]. Cross-linked smaller ones. The erosion of matrices was also polyanhydrides were synthesized from monomers strongly related to their geometry and was highest such as SA, CPP or CPH after conversion to mixed for small matrices. These results indicate that the anhydrides with methacrylic acid. The obtained geometry can indeed affect the progress of degra- methacrylated sebacic acid (MSA), methacrylated dation and erosion. This was also supported by CPP (MCPP) and methacrylated CPH (MCPH) were release experiments with brilliant blue-loaded 924 A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931 p(CPP-SA) 20:80 matrix discs [30]. Erosion models erosion character. In drug delivery applications that confirmed that the changes in release that were means that we are interested in erosion-controlled obtained when the height of 6-mm diameter cylin- release systems. This has sometimes, but not always ders was increased from 1.5 to 2 mm were due to been successful. Usually there are several mecha- surface erosion. The example also illustrates, how- nisms by which the release of a drug from a polymer ever, that taking advantage of device geometry for can be controlled, such as diffusion, polymer swell- modulating erosion and drug release may lead to ing, erosion or the dissolution velocity of the drug. If large devices that may not be useful for biomedical we want to control the release of a drug exclusively applications any more. by one of these mechanisms, which usually compete with one another, we have to make sure that it is the fastest process [80]. From the erosion mechanisms 3 .2.2. Microparticles outlined above, it is obvious that polyanhydrides are Obviously it is much harder to investigate micro- ideal candidates for erosion-controlled release. There spheres, especially with respect to their microstruc- have been many reports that polyanhydride erosion ture. The erosion of microspheres is faster than the and the resulting drug release kinetics are identical. erosion of solid matrices. Microspheres made of For a surface eroding polymer this can mean that we p(FAD-SA) 8:92, p(FAD-SA) 25:75 and p(FAD-SA) have linear release kinetics which was, for example, 44:56 with average diameters below 100 mm re- found for drug release from p(CPP), p(CPP-SA) leased SA in approximately 100 h to 100% in vitro 20:80 matrix discs [54]. That the nature of the [78]. This suggests that the polymer is completely substance to be released and the way by which degraded after that time and that the release of acid erosion is measured has a major impact on this orange that persisted for 400 h was mainly due to the correlation becomes obvious when examining some semisolid erosion zone formed by FAD monomer. data for microspheres prepared from p(CPP-SA) That polymer degradation and eventually also ero- 20:80 [74]. While the more hydrophilic acid orange sion depends on the microsphere size was demon- exhibited a burst release, the more lipophilic p- strated with some early work on p(CPP-SA) 21:79. nitroaniline correlated with CPP release, which was Depending on the average microsphere size (between taken as a measure for erosion. From the erosion 50 and 1100 mm), the microspheres lost between behavior of p(CPP-SA) matrices outlined above one approximately 90 and 75% of CPP when eroded in can assume that the release of CPP lags significantly vitro [74], which is indicative of a similar degree of behind erosion. The release of lipophilic drugs erosion. When p(CPP-SA) 20:80 and p(FA-SA) seems, therefore, to be correlated to dissolution 20:80 microspheres with a size of less than 10 mm rather than erosion. To put polyanhydrides to the test were manufactured by spray drying they degraded on erosion-controlled release it seems, therefore, far almost completely within 18 and 5 h, respectively better to use hydrophilic model compounds [30,81]. [75]. These results illustrate that polyanhydride These model substances are not retained inside a microspheres are very delicate systems that degrade polyanhydride by their hydrophobic character, but and most likely erode very fast, which can make it are released immediately upon erosion [82,83]. hard to use them for long term drug delivery A sensitive issue is the incorporation of proteins applications. This is most likely also one of the and peptides into polyanhydrides. Although there reasons why there have been few reports on poly- have been numerous reports on the incorporation and anhydride nanoparticles [79]. release of proteins and peptides into polyanhydrides [84–86], the polymers may lead, similar to less reactive PLA and PLGA [87], to an acylation of 4 . Erosion-controlled drug release from nucleophiles such as primary amines or hydroxy polyanhydrides groups [88]. This has to be carefully considered when proteins, peptides or drug-carrying amine or One of the goals that we have in mind when using hydroxyl functions are to be released from poly- polyanhydrides is taking advantage of their surface anhydrides. A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931 925

5 . Polyanhydride erosion modeling sdsdsdD 2 2Kt 2 1 2 D 2 2Kt L 2 2Kt f 5 ]]]]]]]]]000] (2) D 2 1 2DL The intriguing advantage of polyanhydrides over 000 most other degradable polymers is their clean-cut Similar approaches were used later taking advan- erosion behavior. As shown above, the erosion of tage of the individual erosion front velocities that these materials is characterized by a linear pro- were determined for polyanhydride matrix cylinders. gression of erosion fronts associated with a strong This allowed to come up with simple predictions on correlation between erosion and drug release for a the release kinetics of drugs [81]. considerable number of drugs. This spurred the interest in the development of theoretical models, 5 .2. Monte Carlo-based models that allow to describe and predict the erosion behavior of polyanhydride matrices [89]. A different approach to polyanhydride erosion modeling was proposed in the early 1990s [92]. The 5 .1. Empirical models concept offers the advantage, that the degradation of the polymer was modeled as a random event that The first models that were developed were based obeyed first-order reaction kinetics. Rather than on the assumption of a linear moving erosion front. describing the degradation of individual bonds, the They were empirical in a sense that they did not degradation of all bonds contained in a small volume relate the erosion behavior of the polymer to measur- of polymer was described. Similar approaches were able parameters with an exactly defined physical used by Zygourakis and co-workers for modeling significance. Hopfenberg, for example, derived a erosion-controlled drug release from eroding compo- general equation for describing erosion-controlled site devices [93,94]. Polymer cross-sections were drug release that may also apply to describing the covered with two-dimensional grids by which the erosion polyanhydride spheres, cylinders and slabs polymer matrix was divided into a multitude of [90]: pixels, each representing a small volume of polymer. To account for the existence of crystalline and Mktn ]t 5 1 2SD1 2 ]0 (1) amorphous areas inside most polyanhydrides, these Mca` 0 pixels were randomly assigned one of these two qualities so that the relative number of pixels desig- where Mt and M` are the polymer mass at time t and nated as crystalline were in agreement with the at infinite time, respectively, c0 a uniform initial drug crystallinity of the polymer to be modeled. The ‘life concentration or in the case of erosion a ‘polymer expectancy’ of a pixel after contact with water was concentration’, a is the radius of a cylinder or sphere sampled at random from first-order Erlang distribu- or the half-thickness of a slab and n is a ‘shape tions: factor’ (n53 for spheres, n52 for cylinders and n51 for slabs). According to Hopfenberg’s model, e(t) 5 l e2lt (3) only slabs erode with zero-order erosion kinetics. Another model of heterogeneous erosion that may where l is a degradation rate constant that is be used to describe erosion was developed by different for amorphous and crystalline polymer. e(t) Cooney [91]. Cooney describes, like Hopfenberg, the is the probability that a polymer pixel degrades at erosion of polymer like a ‘dissolution’ process but time t [95]. Applying direct Monte Carlo sampling assumes that there is an additional step involved, techniques [96], values for t can be obtained at namely the release through an adjacent stationary random so that all values are distributed according to solvent layer into the erosion medium. Cylindrical Eq. (3). By applying this procedure to the two- polymer matrices with an initial length L0 and initial dimensional grids, it is possible to simulate the diameter D0 erode according to Eq. (2) in which f erosion of polyanhydrides (Fig. 9). With these simu- designates the fractional ‘dissolution’ at time t lations it is possible to model the erosion profile or (relative to t50) and K is a rate constant. the porosity of polyanhydride matrices during ero- 926 A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931

closer look at the microstructure of eroded matrices shows that this cannot be unequivocally true. Fig. 10a shows a pictures of the erosion fronts inside a p(CPP-SA) 20:80 matrix. Arrows identify the erosion fronts sharply separating eroded from non- eroded polymer. A look at the front at higher resolution shows that it is not a straight line any more. Rather than being well deﬁned, it has the character of a transition zone in which the polymer changes from eroded to non-eroded within 5–20 mm. Recently developed models can shed some light on the consequences of this uncertainty [97]. It was

Fig. 9. Two-dimensional simulation of p(CPP-SA) 20:80 erosion. Reproduced with permission from Ref. [30]. sion. These data can then be used to ﬁt the model to experimental data, which yields the erosion rate of crystalline and amorphous areas in p(CPP-SA) [92]. Originally the model was limited to two-dimensional simulations; however, it was also possible to expand it to three dimensions by assuming rotation symmetry [30]. This modiﬁcation allowed to simulate the erosion of cylinders and concomitantly the release of drugs from such matrices. In combination with related models for bulk eroding polymers [12], it is also possible to simulate the erosion of compo- site devices that were developed for the manufacture of programmable release devices [83].

6 . Why polyanhydrides undergo surface erosion Fig. 10. Erosion front inside a p(CPP-SA) 20:80 matrix at two different magnifications (reproduced with permission from Ref. So far it has been widely accepted that poly- [10]): (a) survey on a cylinder cross-section (the arrow indicates anhydrides are surface eroding polymers. This is the position of the erosion front), (b) erosion front at higher certainly true for macroscopic devices. However, a magnification. A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931 927 assumed that a polymer matrix erodes according to a considered in design strategies involving poly- surface erosion mechanism if the velocity of water anhydrides. Furthermore the theoretical value of ´ is uptake, as described by diffusion theory, is lower well in agreement with experimental findings regard- than degradation rate of the polymer as described by ing the width of the erosion front. Monte Carlo models [12,92]. A dimensionless ‘erosion number’ ´, which is the ratio of both processes indicates the mode of erosion: 7 . Summary and outlook

2 x lp Polyanhydrides are materials that, based on their KL chemical nature degrade rapidly in a an aqueous ´ 5 ]]]]]]]]]]]]]] ] ] (4) M environment. Passive hydrolysis seems, thereby, to 4D ln[ x ] 2 ln 3 ]]]n ] effS KL FGD be the most signiﬁcant mechanism of polymer bond NA(N 2 1)r œ cleavage. Enzymatic degradation mechanisms seem

where Deff is the effective diffusivity of water to be of minor or no importance for polyanhydrides inside the polymer, kxl the device dimension, l the investigated so far. When the degradation of poly- degradation rate of the polymer, Mn the number anhydride matrices is followed by monitoring poly- average polymer molecular weight, N the average mer molecular weight, usually an exponential decay degree of polymerization, NA Avogadro’s number of molecular weight over time is observed. and r the polymer density. For ´51, the erosion Erosion in contrast is more complicated. Crys- mechanism is not defined and a critical device talline polymer areas were found to be more erosion dimension Lcritical can be calculated. If a matrix is resistant than amorphous ones. In many poly- larger than Lcritical it will undergo surface erosion, if anhydrides such as p(CPP-SA) and p(FAD-SA), not it will be bulk eroding. Lcritical values for erosion fronts moved from the surface of polymer polymers were estimated based on literature data. matrices to their center separating eroded from non- Polyanhydrides were estimated to be surface eroding eroded polymer. These erosion zones may either be 24 down to a size of approximately Lcritical510 m, porous (Fig. 11a) such as in the case of p(CPP-SA), while poly(a-hydroxy acids) matrices need to be or semisolid (Fig. 11b) as was observed for p(FAD- 21 larger than Lcritical510 m to loose their bulk SA). Inside the erosion zone, the pH values drop as erosion properties. To support this theoretical find- determined by EPR to values close to the pKa of the ings it was shown experimentally that poly(a-hy- monomers. Some monomers such as SA may pre- droxy acid) matrices, which are considered classical cipitate under these conditions. bulk eroding materials, can also undergo surface The erosion zone is of utmost importance for the erosion [97]. Although this model yields only a release of compounds from polyanhydrides as it may crude estimate for ´, it suggests that there is certainly be a significant diffusion barrier to be overcome, as a size limit to surface erosion, a fact that needs to be in the case of p(FAD-SA). The environment inside

Fig. 11. Potential erosion mechanisms for polyanhydrides. (a) Erosion under the formation of porous erosion zones such as in p(CPP-SA). (b) Erosion under the formation of semisolid erosion zones such as in p(FAD-SA). 928 A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931 the erosion zone may be of signiﬁcance for the degradable polymers, Proc. Natl. Acad. Sci. USA 90 (1993) stability of drugs. It can stabilize compounds such as 552–556. BCNU which has a stability optimum at low pH [8] D.F. Williams, E. Mort, Enzyme-accelerated hydrolysis of polyglycolic acid, J. Bioeng. 1 (1977) 231–238. [98], it can, however, also put the stability of [9] A.J. Domb, R. Nudelman, In vivo and in vitro elimination of proteins and peptide at risk due to the potential of aliphatic polyanhydrides, Biomaterials 16 (1995) 319–323. anhydrides to acylate nucleophiles. [10] A. Gopferich, R. Langer, The inﬂuence of microstructure and Erosion models helped to describe and understand monomer properties on the erosion mechanism of a class of the erosion behavior of polyanhydrides. Monte Carlo polyanhydrides, J. Polym. Sci., Part A: Polym. Chem. 31 (1993) 2445–2458. models that simulate erosion based on the erosion of [11] M.P. Wu, J.A. Tamada, H. Brem, R. Langer, In vivo versus small individual polymer areas after their contact in vitro degradation of controlled release polymers for with water allowed to explain the formation of intracranial surgical therapy, J. Biomed. Mater. Res. 28 erosion zones and erosion fronts. Models that com- (1994) 387–395. pare the velocity of water diffusion into a poly- [12] A. Goepferich, Polymer bulk erosion, Macromolecules 30 (1997) 2598–2604. anhydride matrix to the velocity of erosion allow to [13] H. Akbari, A. D’Emanuele, D. Attwood, Effect of geometry address why polyanhydrides undergo surface ero- on the erosion characteristics of polyanhydride matrixes, Int. sion. It seems that surface erosion is a characteristic J. Pharm. 160 (1998) 83–89. that is strongly linked to the dimensions of a device [14] A. D’Emanuele, J. Hill, J.A. Tamada, A.J. Domb, R. Langer, and that, below a critical size limit, this property is Molecular weight changes in polymer erosion, Pharm. Res. 9 lost. We should, therefore, examine, polyanhydride (1992) 1279–1283. [15] A.J. Domb, C.F. Gallardo, R. Langer, Poly(anhydrides). 3. devices very carefully when they are smaller than Poly(anhydrides) based on aliphatic-aromatic diacids, Macro- 100 mm. molecules 22 (1989) 3200–3204. In summary polyanhydrides have unique degra- [16] F. Heatley, M. Humadi, R.V. Law, A. D’Emanuele, Erosion of a 1,3-bis( p-carboxyphenoxy)propane-sebacic acid poly- dation and erosion properties, that make them pre- 113 cious materials for numerous medical, biomedical (anhydride) copolymer by water vapor studied by H and C solid-state NMR spectroscopy, Macromolecules 31 (1998) and pharmaceutical applications. If we use them with 3832–3838. the necessary circumspection they are ideal for all [17] L. Shieh, J. Tamada, I. Chen, J. Pang, A. Domb, R. Langer, those applications in which we need degradable Erosion of a new family of biodegradable polyanhydrides, J. polymers that allow for a perfect erosion control. Biomed. Mater. Res. 28 (1994) 1465–1475. [18] E. Ron, E. Mathiowitz, G. Mathiowitz, A. Domb, R. Langer, NMR characterization of erodible copolymers, Macromole- cules 24 (1991) 2278–2282. [19] Z. Grubisic, P. Rempp, H. Benoit, A universal calibration for R eferences gel permeation chromatography, J. Polym. Sci., Part B: Polym. Phys. 34 (1996) 1707–1714. [1] A. Gopferich, R. Gref, Y. Minamitake, L. Shieh, M.J. [20] C.A. Santos, B.D. Freedman, K.J. Leach, D.L. Press, M. Alonso, Y. Tabata, R. Langer, Drug delivery from bioerod- Scarpulla, E. Mathiowitz, Poly(fumaric-co-sebacic anhy- ible polymers: systemic and intravenous administration, ACS dride) A degradation study as evaluated by FTIR, DSC, GPC Symp. Ser. 567 (1994) 242–277. and X-ray diffraction, J. Control. Release 60 (1999) 11–22. [2] R. Langer, Drug delivery and targeting, Nature (London) 392 [21] D.L. McCann, F. Heatley, A. D’Emanuele, Characterization (1998) 5–10. of chemical structure and morphology of eroding poly- [3] R. Langer, J.P. Vacanti, Tissue engineering, Science anhydride copolymers by liquid-state and solid-state 1H (Washington, DC) 260 (1993) 920–926. NMR, Polymer 40 (1998) 2151–2162. [4] R.T. Bartus, M.A. Tracy, D.F. Emerich, S.E. Zale, Sustained [22] M. Sandor, N.A. Bailey, E. Mathiowitz, Characterization of delivery of proteins for novel therapeutic products, Science polyanhydride microsphere degradation by DSC, Polymer 43 (Washington, DC) 281 (1998) 1161–1162. (2001) 279–288. [5] J.B. Kohn, S. Brocchini, K. James, V. Tangpasuthadol, [23] K.M. Shakesheff, M.C. Davies, C.J. Roberts, S.J.B. Tendler, Copolymer libraries, their production and monomers there- A.G. Shard, A. Domb, In situ atomic force microscopy for, WO9952962. imaging of polymer degradation in an aqueous environment, [6] H.B. Rosen, J. Chang, G.E. Wnek, R.J. Linhardt, R. Langer, Langmuir 10 (1994) 4417–4419. Bioerodible polyanhydrides for controlled drug delivery, [24] K.M. Shakesheff, X. Chen, M.C. Davies, A. Domb, C.J. Biomaterials 4 (1983) 131–133. Roberts, S.J.B. Tendler, P.M. Williams, Relating the phase [7] J.A. Tamada, R. Langer, Erosion kinetics of hydrolytically morphology of a biodegradable polymer blend to erosion A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931 929

kinetics using simultaneous in situ atomic force microscopy [39] L. Erdmann, B. Macedo, K.E. Uhrich, Degradable poly- and surface plasmon resonance analysis, Langmuir 11 (1995) (anhydride ester) implants: effects of localized salicylic acid 3921–3927. release on bone, Biomaterials 21 (2000) 2507–2512. [25] S.L. McGurk, K.M. Shakesheff, M.C. Davies, A. Domb, C.J. [40] T.J. Anastasiou, K.E. Uhrich, Synthesis of novel, degradable Roberts, S.J.B. Tendler, P.M. Williams, Rapid screening of polyanhydrides containing para-aminosalicylic acid as drug biodegradable polymers by AFM and SPR, Proc. Int. Symp. delivery devices for tuberculosis treatment, Abstr. Pap. Am. Control. Release Bioact. Mater. 24 (1997) 947–948. Chem. Soc. 220 (2000) POLY272. [26] E. Mathiowitz, E. Ron, G. Mathiowitz, C. Amato, R. Langer, [41] H.L. Jiang, K.J. Zhu, Synthesis, characterization and in vitro Morphological characterization of bioerodible polymers. 1. degradation of a new family of alternate poly(ester-anhy- Crystallinity of polyanhydride copolymers, Macromolecules drides) based on aliphatic and aromatic diacids, Biomaterials 23 (1990) 3212–3218. 22 (2000) 211–218. [27] K. Maeder, Y. Cremmilleux, A.J. Domb, J.F. Dunn, H.M. [42] H.L. Jiang, K.J. Zhu, Preparation, characterization and Swartz, In vitro/in vivo comparison of drug release and degradation characteristics of polyanhydrides containing polymer erosion from biodegradable P(FAD-SA) poly- poly(ethylene glycol), Polym. Int. 48 (1999) 47–52. anhydrides—a noninvasive approach by the combined use of [43] D. Chickering III, E. Jacob, E. Mathiowitz, Poly(fumaric-co- electron paramagnetic resonance spectroscopy and nuclear sebacic) anhydride microspheres as oral drug delivery sys- magnetic resonance imaging, Pharm. Res. 14 (1997) 820– tems, Biotechnol. Bioeng. 52 (1996) 96–101. 826. [28] K. Mader, S. Nitschke, R. Stosser, H.H. Borchert, A. Domb, [44] A.C. Albertsson, Y. Liu, Comparison between physical Nondestructive and localized assessment of acidic microen- blending and copolymerization of poly(trimethylene carbon- vironments inside biodegradable polyanhydrides by spectral ate) and poly(adipic anhydride) with special regard to spatial electron paramagnetic resonance imaging, Polymer 38 compatibility, morphology and degradation, J. Macromol. (1997) 4785–4794. Sci., Pure Appl. Chem. A34 (1997) 1457–1482. [29] J. Tamada, M. Sawka, R. Langer, Probing the nature of the [45] K.W. Leong, B.C. Brott, R. Langer, Bioerodible poly- erosion of polyanhydrides, Proc. Int. Symp. Control. Release anhydrides as drug-carrier matrices. I: Characterization, Bioact. Mater. 18 (1991) 327–328. degradation, and release characteristics, J. Biomed. Mater. [30] A. Goepferich, R. Langer, Modeling of polymer erosion in Res. 19 (1985) 941–955. three dimensions: rotationally symmetric devices, AIChE J. [46] C. Bedell, M. Deng, T.J. Anastasiou, K.E. Uhrich, Process- 41 (1995) 2292–2299. ing and hydrolytic degradation of aromatic, ortho-substituted [31] C.T. Laurencin, S.E.M. Ibim, R.S. Langer, Poly(anhydrides), polyanhydrides, J. Appl. Polym. Sci. 80 (2001) 32–38. in: J.O. Hollinger (Ed.), Biomed. Appl. Synth. Biodegrad. [47] K.W. Leong, V. Simonte, R. Langer, Synthesis of poly- Polym, CRC Press, Boca Raton, FL, 1995, pp. 59–102. anhydrides: melt-polycondensation, dehydrochlorination, and [32] K. Leong, A. Domb, E. Ron, R. Langer, Polyanhydrides, dehydrative coupling, Macromolecules 20 (1987) 705–712. Encyclopedia of Polymer Science and Engineering, John [48] A.J. Domb, R. Langer, Polyanhydrides. I. Preparation of Wiley & Sons Inc. Suppl. Volume (1989). high molecular weight polyanhydrides, J. Polym. Sci., Part [33] J. Tamada, R. Langer, The development of polyanhydrides A: Polym. Chem. 25 (1987) 3373–3386. for drug delivery applications, J. Biomater. Sci., Polym. Ed. [49] H. Akbari, A. D’Emanuele, D. Attwood, Effect of fabrication 3 (1992) 315–353. technique on the erosion characteristics of polyanhydride [34] J. Hanes, M. Chiba, R. Langer, Degradation of porous matrixes, Pharm. Dev. Technol. 3 (1998) 251–259. poly(anhydride-co-imide) microspheres and implications for [50] A. D’Emanuele, J. Kost, J.L. Hill, R. Langer, An inves- controlled macromolecule delivery, Biomaterials 19 (1998) tigation of the effects of ultrasound on degradable poly- 163–172. anhydride matrices, Macromolecules 25 (1992) 511–515. [35] J. Hanes, M. Chiba, R. Langer, Synthesis and characteriza- [51] K. Maeder, G. Bacic, A. Domb, O. Elmalak, R. Langer, tion of degradable anhydride-co-imide terpolymers contain- H.M. Swartz, Noninvasive in vivo monitoring of drug ing trimellitylimido-L-tyrosine: novel polymers for drug release and polymer erosion from biodegradable polymers by delivery, Macromolecules 29 (1996) 5279–5287. EPR spectroscopy and NMR imaging, J. Pharm. Sci. 86 [36] K.E. Uhrich, S.E.M. Ibim, D.R. Larrier, R. Langer, C.T. (1997) 126–134. Laurencin, Chemical changes during in vivo degradation of [52] H. Brem, Polymers to treat brain tumors, Biomaterials 11 poly(anhydride-imide) matrixes, Biomaterials 19 (1998) (1990) 699–701. 2045–2050. [53] H. Brem, A. Kader, J.I. Epstein, R.J. Tamargo, A. Domb, R. [37] K.E. Uhrich, D.R. Larrier, C.T. Laurencin, R. Langer, In Langer, K.W. Leong, Biocompatibility of a biodegradable, vitro degradation characteristics of poly(anhydride-imides) controlled-release polymer in the rabbit brain, Sel. Cancer containing pyromellitylimidoalanine, J. Polym. Sci., Part A: Ther. 5 (1989) 55–65. Polym. Chem. 34 (1996) 1261–1269. [54] M. Chasin, A. Domb, E. Ron, E. Mathiowitz, R. Langer, K. [38] T.J. Anastasiou, M.L. Beaton, K.E. Uhrich, Syntheses of Leong, C. Laurencin, H. Brem, S. Grossman, Polyanhydrides aminosalicylate-based polyanhydride prodrugs: esters, as drug delivery systems, Drugs Pharm. Sci. 45 (1990) amides, and azos, Polym. Prepr. (Am. Chem. Soc., Div. 43–70. Polym. Chem.) 42 (2001) 121–122. [55] H. Brem, P. Piantadosi et al., Placebo/controlled trial of 930 A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931

safety and efﬁcacy of intraoperative controlled delivery by microspheres: 3. Morphology and characterization of sys- biodegradable polymers of chemotherapy for recurrent tems made by solvent removal, Polymer 31 (1990) 547–555. gliomas, Lancet 345 (1995) 1008–1012. [74] E. Mathiowitz, R. Langer, Polyanhydride microspheres as [56] E.S. Park, M. Maniar, J. Shah, Water uptake in to poly- drug carriers, I. Hot-melt microencapsulation, J. Control. anhydride devices: kinetics of uptake and effects of model Release 5 (1987) 13–22. compounds incorporated, and device geometry on water [75] E. Mathiowitz, H. Bernstein, S. Giannos, P. Dor, T. Turek, R. uptake, J. Control. Release 40 (1996) 55–65. Langer, Polyanhydride microspheres. IV. Morphology and [57] A. Goepferich, L. Schedl, R. Langer, The precipitation of characterization of systems made by spray drying, J. Appl. monomers during the erosion of a class polyanhydrides, Polym. Sci. 45 (1992) 125–134. Polymer 37 (1996) 3861–3869. [76] E. Mathiowitz, R. Langer, Polyanhydride microspheres as [58] A.J. Domb, M. Maniar, Absorbable biopolymers derived drug delivery systems, Microcapsules Nanopart. Med. from dimer fatty acids, J. Polym. Sci., Part A: Polym. Chem. Pharm., pp. 99–123. 31 (1993) 1275–1285. [77] K.J. Pekarek, M.J. Dyrud, K. Ferrer, Y.S. Jong, E. [59] L. Shieh, J. Tamada, Y. Tabata, A. Domb, R. Langer, Drug Mathiowitz, In vitro and in vivo degradation of double- release from a new family of biodegradable polyanhydrides, walled polymer microspheres, J. Control. Release 40 (1996) J. Control. Release 29 (1994) 73–82. 169–178. [60] L. Shieh, A. Goepferich, R. Langer, The erosion properties [78] Y. Tabata, R. Langer, Polyanhydride microspheres that of polyanhydrides, Mater. Res. Soc. Symp. Proc. 331 (1994) display near-constant release of water-soluble model drug 85–90. compounds, Pharm. Res. 10 (1993) 391–399. [61] A.J. Domb, R. Nudelman, Biodegradable polymers derived [79] J. Fu, C. Wu, Laser light scattering study of the degradation from natural fatty acids, J. Polym. Sci., Part A: Polym. of poly(sebacic anhydride) nanoparticles, J. Polym. Sci., Part Chem. 33 (1995) 717–725. B: Polym. Phys. 39 (2001) 703–708. [62] D. Teomim, A. Nyska, A.J. Domb, Ricinoleic acid-based [80] A. Goepferich, Polymer degradation and erosion. Mecha- biopolymers, J. Biomed. Mater. Res. 45 (1999) 258–267. nisms and applications, Eur. J. Pharm. Biopharm. 42 (1996) [63] D. Teomim, A.J. Domb, Fatty acid terminated poly- 1–11. anhydrides, J. Polym. Sci., Part A: Polym. Chem. 37 (1999) [81] A. Goepferich, D. Karydas, R. Langer, Predicting drug 3337–3344. release from cylindric polyanhydride matrix discs, Eur. J. [64] D. Teomim, A.J. Domb, Nonlinear fatty acid terminated Pharm. Biopharm. 41 (1995) 81–87. polyanhydrides, Biomacromolecules 2 (2001) 37–44. [82] A. Goepferich, Erosion-controlled pulsatile release, Proc. Int. [65] A.J. Domb, E. Mathiowitz, E. Ron, S. Giannos, R. Langer, Symp. Control. Release Bioact. Mater. 23 (1996) 220–221. Polyanhydrides. IV. Unsaturated and crosslinked poly- [83] A. Goepferich, Bioerodible implants with programmable anhydrides, J. Polym. Sci., Part A: Polym. Chem. 29 (1991) drug release, J. Control. Release 44 (1997) 271–281. 571–579. [84] K.L. Smith, M.E. Schimpf, K.E. Thompson, Bioerodible [66] K.S. Anseth, D.J. Quick, Polymerizations of multifunctional polymers for delivery of macromolecules, Adv. Drug Deliv. anhydride monomers to form highly crosslinked degradable Rev. 4 (1990) 343–357. networks, Macromol. Rapid Commun. 22 (2001) 564–572. [85] E. Ron, T. Turek, E. Mathiowitz, M. Chasin, M. Hageman, R. Langer, Controlled release of polypeptides from poly- [67] K.S. Anseth, V.R. Shastri, R. Langer, Photopolymerizable anhydrides, Proc. Natl. Acad. Sci. USA 90 (1993) 4176– degradable polyanhydrides with osteocompatibility, Nat. 4180. Biotechnol. 17 (1999) 156–159. [86] Y. Tabata, S. Gutta, R. Langer, Controlled delivery systems [68] A.K. Burkoth, K.S. Anseth, A review of photocrosslinked for proteins using polyanhydride microspheres, Pharm. Res. polyanhydrides: in situ forming degradable networks, 10 (1993) 487–496. Biomaterials 21 (2000) 2395–2404. [87] A. Lucke, J. Kiermaier, A. Goepferich, Peptide acylation by [69] A.K. Burkoth, K.S. Anseth, Photocurable polyanhydrides poly(a-hydroxy esters), Pharm. Res. 19 (2002) 175–181. engineered for orthopaedic applications, Abstr. Pap. Am. [88] A.J. Domb, L. Turovsky, R. Nudelman, Chemical interac- Chem. Soc. 220 (2000) PMSE288. tions between drugs containing reactive amines with hydro- [70] A.K. Burkoth, J. Burdick, K.S. Anseth, Surface and bulk lyzable insoluble biopolymers in aqueous solutions, Pharm. modiﬁcations to photocrosslinked polyanhydrides to control Res. 11 (1994) 865–868. degradation behavior, J. Biomed. Mater. Res. 51 (2000) [89] J. Siepmann, A. Gopferich, Mathematical modeling of 352–359. bioerodible, polymeric drug delivery systems, Adv. Drug [71] D.S. Muggli, A.K. Burkoth, K.S. Anseth, Crosslinked poly- Deliv. Rev. 48 (2001) 229–247. anhydrides for use in orthopedic applications: degradation [90] H.B. Hopfenberg, Controlled release from erodible slabs, behavior and mechanics, J. Biomed. Mater. Res. 46 (1999) cylinders and spheres, in: D.R. Paul, F.W. Harris (Eds.), 271–278. Controlled Release Polymeric Formulations, Vol. 33, Ameri- [72] K.J. Leach, E. Mathiowitz, Degradation of double-walled can Chemical Society, Washington, DC, 1976, pp. 26–32. polymer microspheres of PLLA and P(CPP:SA)20:80. I. In [91] D.O. Cooney, Effect of geometry on the dissolution of vitro degradation, Biomaterials 19 (1998) 1973–1980. pharmaceutical tablets and other solids. Surface detachment [73] E. Mathiowitz, C. Amato, Ph. Dor, R. Langer, Polyanhydride kinetics controlling, AIChE J. 18 (1972) 446–449. A. Gopferich¨ , J. Tessmar / Advanced Drug Delivery Reviews 54 (2002) 911–931 931

[92] A. Gopferich, R. Langer, Modeling of polymer erosion, [95] A.W. Drake, in: Fundamentals of Applied Probability Macromolecules 26 (1993) 4105–4112. Theory, McGraw-Hill, New York, 1988. [93] K. Zygourakis, Development and temporal evolution of [96] M.H. Kalos, P.A. Whitlock, in: Monte Carlo Methods, Wiley, erosion fronts in bioerodible controlled release devices, New York, 1986. Chem. Eng. Sci. 45 (1990) 2359–2366. [97] F. von Burkersroda, L. Schedl, A. Goepferich, Why degrad- [94] K. Zygourakis, P.A. Markenscoff, Computer-aided design of able polymers undergo surface or bulk erosion, Biomaterials bioerodible devices with optimal release characteristics: a 23 (2002) 4221–4231. cellular automata approach, Biomaterials 17 (1996) 125– [98] P.A. Laskar, J.W. Ayres, Degradation of carmustine in 135. aqueous media, J. Pharm. Sci. 66 (1977) 1073–1076.